GROMACS version: 2024.4
GROMACS modification: No
I used the following command to compile gromacs in a virtual machine from Azure.
cmake … -DCMAKE_C_COMPILER=gcc-10
-DCMAKE_CXX_COMPILER=g+±10
-DGMX_MPI=on
-DGMX_GPU=CUDA
-DGMX_BUILD_OWN_FFTW=ON
-DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda
The drivers that I have are 560.35.05 and nvcc-CUDA Version: 12.6. The VM has 8 vCPUs and 1 Nvidia Tesla T4 GPU (16GB).
Start 31: GpuUtilsUnitTests
31/89 Test #31: GpuUtilsUnitTests … Passed 1.04 sec
Start 73: MdrunMpiTests
73/89 Test #73: MdrunMpiTests … Passed 8.96 sec
Start 74: MdrunMultiSimTests
74/89 Test #74: MdrunMultiSimTests … Passed 11.52 sec
Start 75: MdrunMultiSimReplexTests
75/89 Test #75: MdrunMultiSimReplexTests … Passed 5.60 sec
Start 76: MdrunMultiSimReplexEquivalenceTests
76/89 Test #76: MdrunMultiSimReplexEquivalenceTests … Passed 14.36 sec
Start 77: MdrunMpi1RankPmeTests
77/89 Test #77: MdrunMpi1RankPmeTests … Passed 4.11 sec
Start 78: MdrunMpi2RankPmeTests
78/89 Test #78: MdrunMpi2RankPmeTests … Passed 10.43 sec
Label Time Summary:
GTest = 1070.93 secproc (87 tests)
IntegrationTest = 634.58 secproc (30 tests)
MpiTest = 805.20 secproc (23 tests)
QuickGpuTest = 259.04 secproc (20 tests)
SlowGpuTest = 730.61 secproc (14 tests)
SlowTest = 382.64 secproc (13 tests)
UnitTest = 53.71 sec*proc (44 tests)
All the test performed by make check were passed even those regarding GPU use, though when I check for gmx --version the GPU support appears to be disabled.
GROMACS: gmx, version 2021.4-Ubuntu-2021.4-2
Executable: /usr/bin/gmx
Data prefix: /usr
Working dir: /home/azureuser/gromacs-2024.4/build
Command line:
gmx --version
GROMACS version: 2021.4-Ubuntu-2021.4-2
Precision: mixed
Memory model: 64 bit
MPI library: thread_mpi
OpenMP support: enabled (GMX_OPENMP_MAX_THREADS = 64)
GPU support: disabled
SIMD instructions: SSE4.1
FFT library: fftw-3.3.8-sse2-avx
RDTSCP usage: enabled
TNG support: enabled
Hwloc support: hwloc-2.5.0
Tracing support: disabled
C compiler: /usr/bin/cc GNU 11.2.0
C compiler flags: -msse4.1 -Wno-missing-field-initializers -fexcess-precision=fast -funroll-all-loops -O3 -DNDEBUG
C++ compiler: /usr/bin/c++ GNU 11.2.0
C++ compiler flags: -msse4.1 -Wno-missing-field-initializers -fexcess-precision=fast -funroll-all-loops -fopenmp -O3 -DNDEBUG
I am trying to provide all information that I think might be related to the use of the GPUs, hope that helps. One thing that has catch my eye is some test that do not respond when the nvcc is trying to make some checks.
– Checking if nvcc accepts flags --generate-code=arch=compute_35,code=sm_35
– Checking if nvcc accepts flags --generate-code=arch=compute_35,code=sm_35 - No
– Checking if nvcc accepts flags --generate-code=arch=compute_37,code=sm_37
– Checking if nvcc accepts flags --generate-code=arch=compute_37,code=sm_37 - No
These appear during the cmake and make processes.
Thanks in advance, hope you have an idea of what the issue is and I was clear enough.